MediaTek and NVIDIA have built a new class of Windows PC processors for personal agents. An agent takes a goal and works through it on its own, without a prompt at every step. That calls for fast compute, large memory, and low power use. NVIDIA RTX Spark delivers this, and MediaTek is central to how.
Its engineering shapes the CPU, memory system, power design, and connectivity inside this RTX processor.
MediaTek’s Role in the RTX Spark Superchip
MediaTek co-designed the Grace CPU with NVIDIA. It offers up to 20 cores and runs the latest Windows applications. For businesses, MediaTek’s CPU and cache design keeps everyday apps responsive and stable.
NVIDIA’s Blackwell RTX GPU adds up to 6,144 cores. Together, the platform reaches up to 1 petaflop of FP4 AI performance. FP4 is a compact number format that speeds up AI math, so large models run faster on the device.
Memory Built for Large AI Models
MediaTek designed a proprietary memory controller. It supports up to 128GB of unified memory and up to 300 GB/s of bandwidth. The CPU and GPU share one pool, so creative projects and AI models can run side by side.
NVIDIA says RTX Spark can run 120-billion-parameter language models locally, with up to 1 million tokens of context. For teams, that means serious AI work on the device, without any reliance on the cloud.
One Chiplet, Slimmer Devices, Longer Battery Life
MediaTek placed the CPU cluster, coherent fabric, memory controller, USB, and PCIe on a single chiplet, working with NVIDIA and TSMC. This lets developers build thin laptops and mini desktops, so employees carry a lighter device without giving up AI performance.
Furthermore, MediaTek also added its low-power architecture and power management ICs, which cut power use even under heavy workloads. MediaTek calls RTX Spark the most power-efficient RTX chipset ever made and says it supports all-day battery life.
RTX Spark Laptops Now, Compact Desktops Next
RTX Spark arrives in two stages. An RTX Spark laptop is available now from ASUS, Dell, HP, Lenovo, Microsoft, and MSI. Launch models include the ASUS ProArt P14 and P16, Dell XPS 16, HP OmniBook Ultra 16 and OmniBook X 14, Lenovo Yoga Pro 9n, Microsoft Surface Laptop Ultra and MSI Prestige N16 Flip AI+.
An RTX Spark desktop follows. It is a compact system built to run personal agents 24/7 at your desk, with room for gaming and creation. Acer has shown an SFF design with up to 128GB, with no launch date yet. Many people call these compact systems an RTX Spark mini PC, and MediaTek’s dense integration is what makes them possible.
Connectivity That Keeps Agents in Sync
MediaTek powers over 2 billion connected devices every year, and RTX Spark applies that experience. It uses MediaTek’s connectivity technologies, from integrated USB and PCIe interfaces to Wi-Fi and Bluetooth. These give network adapters and peripherals fast links to the PC.
Additionally, low-latency wireless also keeps local agents such as NVIDIA NemoClaw in sync with cloud resources.
What Developers and Creators Gain
CUDA runs natively on RTX Spark, so developers can keep their models and tools without rewriting them. They can prototype, fine-tune, and run agents locally, with a path to deploy across more than 150 million NVIDIA RTX AI PCs and cloud GPUs. Local processing keeps data on the device and avoids per-token cloud inference costs.
Creators gain FP4 Tensor Cores, Ray Tracing Cores, and DLSS. NVIDIA says users can edit 8K 4:2:2 video and render 3D scenes up to 90GB, even unplugged.
Conclusion
RTX Spark shows how MediaTek’s strengths in CPU design, memory, power, and connectivity become business value: capable on-device AI, slimmer hardware, and longer working hours. Get ready for a worry-free AI PC experience!
FAQs
What is NVIDIA RTX Spark?
It is a new class of processors for Windows PCs, built by NVIDIA with MediaTek. It targets AI agents, content creation, and gaming in slim laptops and compact desktops.
What does MediaTek’s memory controller add?
It supports up to 128GB of unified memory shared by the CPU and GPU, with up to 300 GB/s of bandwidth. Larger AI models can run on the device.
Does it run existing AI tools?
Yes. CUDA runs natively, so developers can use current models and workflows without rewriting them.
Which brands offer RTX Spark laptops?
ASUS, Dell, HP, Lenovo, Microsoft, and MSI. Examples include the Surface Laptop Ultra and Dell XPS 16.
Top comments (0)